Characterization of Oral Microbiome and Exploration of Potential Biomarkers in Patients with Pancreatic Cancer

Hindawi BioMed Research International Volume 2020, Article ID 4712498, 11 pages https://doi.org/10.1155/2020/4712498

Research Article Characterization of Oral Microbiome and Exploration of Potential Biomarkers in Patients with Pancreatic Cancer

Haiyang Sun,1 Xia Zhao,2 Yanxia Zhou,1 Jun Wang,1 Rui Ma,1 Xi Ren,1 Huaizhi Wang ,3 and Lingyun Zou 1

1Shenzhen Baoan Women’s and Children’s Hospital, Jinan University, 518102 Shenzhen, China 2Bioinformatics Center, Department of Microbiology, Army Medical University, 400038 Chongqing, China 3Institute of Hepatopancreatobiliary Surgery, Chongqing General Hospital, University of Chinese Academy of Sciences, China

Correspondence should be addressed to Huaizhi Wang; [email protected] and Lingyun Zou; [email protected]

Received 26 May 2020; Revised 8 September 2020; Accepted 29 September 2020; Published 2 November 2020

Academic Editor: Koichiro Wada

Copyright © 2020 Haiyang Sun et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Pancreatic cancer (PC) is highly malignant and lacks an effective therapeutic schedule, hence that early diagnosis is of great importance to achieve a good prognosis. Oral bacteria have been proved to be associated with pancreatic cancer, but the specific mechanism has not been comprehensively illustrated. In our study, thirty-seven saliva samples in total were collected with ten from PC patients, seventeen from benign pancreatic disease (BPD) patients, and ten from healthy controls (HC). The oral bacterial community of HC, PC, and BPD groups was profiled by 16S rDNA high-throughput sequencing and bioinformatic methods. As shown by Simpson, Inverse Simpson, Shannon and Heip, oral microbiome diversity of HC, BPD and PC groups is in increasing order with the BPD and PC groups significantly higher than the HC group. Principal coordinate analysis (PCoA) suggested that grouping by PC, BPD and HC was statistically significant. The linear discriminant analysis effect size (LEfSe) identified high concentrations of Fusobacterium periodonticum and low concentrations of Neisseria mucosa as specific risk factors for PC. Furthermore, predicted functions showed changes such as RNA processing and modification as well as the pathway of NOD-like receptor signaling occurred in both PC and HC groups. Conclusively, our findings have confirmed the destruction of oral bacterial community balance among patients with PC and BPD and indicated the potential of Fusobacterium periodonticum and Neisseria mucosa as diagnostic biomarkers of PC.

1. Introduction the regions with the highest microbiome density and the largest number of species in the body [10]. Many studies have Pancreatic cancer (PC) is a common malignant tumor, leading demonstrated that oral bacteria play an important role in the to around 432,242 deaths worldwide in 2018 [1]. Patients in the development of cancer, especially the tumors of oral and gas- early stage of PC have few symptoms, and the disease is com- trointestinal tract [11]. For example, Fusobacterium nuclea- monly diagnosed during intermediate and advanced stages tum and Porphyromonas gingivalis have been frequently for which there is no effective clinical treatment [2]. The five- researched and discussed [12–14]. The oral microbes can year relative survival rate of PC patients was around 9% [3]. communicate with other tissues of the digestive system Therefore, looking for new potential biomarkers will benefit through blood circulation, lymphokines, or biliary conduction PC patient’s early diagnosis greatly. CA19.9 (carbohydrate [15], and oral bacteria were usually associated with cancer of antigen 19.9) was released into blood by PC cells and widely the entire digestive system [16]. Previous 16S sequencing stud- used in clinical diagnosis [4, 5]. However, the testing result ies revealed the crucial role of oral bacteria in the development based on CA19.9, as well as other serological markers such as of pancreatic cancer, and several potential bacterial bio- CEA (carcinoembryonic antigen) are not so reliable [6–8]. markers were identified including Aggregatibacter actinomyce- Lots of diseases were related to the imbalance in the temcomitans, Neisseria elongata, Porphyromonas gingivalis, body’s microbiome [4, 9]. Actually, the oral cavity is one of and Streptococcus mitis [17–19]. It can thus be seen that the 2 BioMed Research International

° community composition of oral microbes provided clues for at -80 C within 1 hour. Samples mixed with blood are early diagnosis, monitoring, prevention, and treatment of PC. not retained. The serological markers are generally secreted into pancreatic juice or blood via pancreatic epithelial tissue cells, 2.2. DNA Isolation, PCR, and Sequencing. The bacterial DNA hence sampling is inconvenient. However, sampling saliva was extracted from the salivary samples using the FastDNA® and test of it are relatively easy. Although several oral bacteria SPIN Kit for Soil (Majorbio, Shanghai, China) following the are found to be related with PC, the association between oral manufacturer’s protocol, and DNAs were quantified via bacteria and PC as well as BPD and the pathogenic mecha- NanoDrop 2000. The V3–V4 region of the bacterial 16S nism were not explored sufficiently [12, 14]. Under such cir- ribosomal RNA (rRNA) gene, which is generally considered cumstances, we carried out this research to discover new the standard target, was PCR-amplified using universal available salivary biomarkers of PC, and to comprehensively primers 338F 5′-ACTCCTACGGGAGGCAGCAG-3′ and explain the potential mechanism of oral microbes in the 806R 5′-GGACTACHVGGGTWTCTAAT-3′. Meanwhile, pathogenesis of PC. Thirty-seven saliva samples were col- an eight-base sequence served as the barcode was added, lected from ten PC patients, seventeen BPD patients, and which was unique to each sample. PCR reactions (20 μL) ten HC subjects. Based on 16S rDNA sequencing results, contained 4 μL of 5x FastPfu Buffer, 2 μL of 2.5 mM dNTPs, we compared the microbial community among the three 0.8 μL of each primer (5 μM), 0.4 μL of TransStart FastPfu groups and demonstrated that microbiome diversity in both DNA Polymerase, 0.2 μL BSA, 10 ng template DNA, and the BPD and PC groups was greatly higher than that in the ddH O. Thermal cycling consisted of the initial denaturation 2° ° ° HC group. LEfSe outputs disclosed that Fusobacterium peri- at 95 C for 2 min, followed by 25 cycles at 95 C for 30 s, 55 C ° ° odonticum and Neisseria mucosa may be the potential bio- for 30 s, 72 C for 45 s and a final extension at 72 C for 10 min. markers of PC. Furtherly, the predicted bacterial gene The construction of high-throughput sequencing libraries functions as well as the enriched pathways showed the signif- and pair-end sequencing was performed using the Illumina icant variance between PC and HC groups concerning RNA MiSeq platform according to the manufacturer’s recommen- processing and modification and enriched pathways such as dations. Finally, all the raw reads were deposited into the the pathway of NOD-like receptor signaling. Our findings NCBI Sequence Read Archive (SRA) database (Bioproject found the association between oral bacteria and PC, and pos- accession Number: SRP237984). sible applications of oral bacteria in diagnostic and treatment strategies for PC. 2.3. Bioinformatic and Statistical Analysis. Firstly, after the raw DNA fragments were generated, we merged the paired- 2. Materials and Methods end reads into a single sequence by means of using FLASH software v.1.2.10 (Fast Length Adjustment of SHort reads) 2.1. Subjects and Sample Collection. A total of ten PC patients, [20]. Then, 16S rRNA OTUs (operational taxonomic units) seventeen BPD patients and ten HC subjects were enrolled in were selected from the combined reads via QIIME toolkit this study. Exclusion criteria included antibiotic treatment v.1.9.1 (Quantitative Insights into Microbial Ecology) [21]. within 8 weeks, oral disease, digestive disease, genetic disease, Note that sequences with 97% identity were assigned to the other cancers and immune system diseases at the time of same OTU and the one with the highest frequencies was sample collection. Both PC and BPD groups were diagnosed selected as the representative sequence. Thirdly, we used by endoscopic ultrasonography and histopathologic the RDP classifier v.2.2 (The Ribosomal Database Project) examination at the First Hospital Affiliated to Army Medical to annotate taxonomic information for each representative University, China. It should be noted that the BPD group sequence [22]. Before performing the subsequent analysis, includes pancreatitis, chronic pancreatitis, and benign pan- we used USEARCH 11 to randomly select an equal number creatic tumors. This study was approved by the Ethics Board of OTU sequences for all samples according to the minimum at the Army Medical University and performed in accor- number of OTU sequences in samples. And next, based on dance with the Helsinki Declaration and Rules of Good the subordinate OTUs of different research groups, the corre- Clinical Practice. Written informed consent upon enrollment sponding alpha diversity was calculated on the indexes that as well as questionnaires addressing past medical history and include Simpson, Heip and Shannon. Note that Simpson lifestyles was obtained from all the volunteers. employed here shows the negative correlation with bacterial In order to avoid the medical intervention that provokes diversity, and other indexes are positively correlated. The sta- an alteration in the oral microbiome, the samples were taken tistical differences of the calculated alpha diversity indexes of from each enrolled subject as close as possible to the time of three groups were analyzed by a nonparametric factorial enrolment. The volunteers were asked not to brush their Kruskal-Wallis sum-rank test (K-W test) and post hoc teeth the night before and the morning of sampling, and Dunn’s multiple comparison test (Dunn’s test) [23]. Beta not to have breakfast. At eight a.m., the volunteers gargled diversity distance between samples was calculated using prin- their mouths with sterile saline before sampling, then held cipal coordinate analysis (PCoA) to explain the phylogenetic the natural, unstimulated saliva in their mouths for about 1 variation based on the weighted UniFrac values. Beta diver- minute before slowly spitting it into the samplers. For each sity comparisons were done using analysis of similarities via volunteer, 10 mL of saliva was collected and transferred to ANNOSIM and NMDS. LEfSe was employed to compare the cryopreservation box immediately, then it was returned the relative abundance of bacterial taxa between the groups, to the laboratory for dispensing and numbering, and frozen and the results were profiled via cladograms and taxonomic BioMed Research International 3 bar charts [24]. Only the bacterial taxa with a linear discrim- SD = 7:088; 3.514, SD = 0:393) and then the PC group inant analysis (LDA) score greater than a certain value and (19.591, SD = 8:166; 3.587, SD = 0:430) (Figure S1). presented statistically significant (P value < 0.05) can be con- As shown in Figure 1(c), the weighted UniFrac analysis sidered the meaningfully different taxa. KEGG pathways and depicted the distance relationship between samples of the COG (clusters of orthologous groups) functions were pre- PC, BPD, and HC groups. ANOSIM analysis showed that dicted by PICRUSt (phylogenetic investigation of communi- the classification of the three groups of samples yielded a ties by reconstruction of unobserved states) based on 16S P value of 0.02 and a r value of 0.12, which indicated that rDNA sequencing data [25]. Correlation study shows that the among-group difference was greater than the intragroup PICRUSt’s prediction accuracy can even reach 0.95, and it difference. To further reveal the significance of grouping, we is recognized as one of the most suitable function prediction performed analysis of NMDS, and the results exhibited that tools [25]. The predicted KEGG and COG results of the stress value of PC, BPD, and HC was 0.11 (less than 0.2). PICRUSt were compared between groups according to the To sum up, the three groups in this study have statistically sequence number of the corresponding results using the significant differences in the bacterial community structure, K-W test and Dunn’s test. Statistical significance was and the grouping of PC and HC is the most explanatory. accepted at P value < 0.05. 3.3. The Oral Bacterial Composition. As shown in Figure 1(d), 3. Results the statistical analysis for bacterial abundance revealed the differences of the oral bacterium community structure 3.1. Clinical Characteristics of the Participants. After applying among the PC, BPD, and HC groups. At the phylum level, rigorous inclusion and exclusion criteria, a total of ten PC the top five bacterial members of the PC, BPD, and HC patients, seventeen BPD patients and ten HC subjects were groups were identical, including Actinobacteria, Bacteroi- finally involved in this study. All participants were in the detes, Firmicutes, Fusobacteria, and Proteobacteria, and their state of oral healthy. Table S1 detailed the clinical proportion in total OTUs were 95.28%, 95.85%, and 98.06%, characteristics of volunteers. There are no significant respectively. However, the relative proportions of dominant differences between these groups in terms of gender, bacteria have changed. Both the dominant bacteria in the smoking and drinking (P value < 0.05). Tumor HC group (45.60%) and the BPD group (29.40%) were Pro- differentiation exhibited that ten PC patients were teobacteria. In contrast, the dominant bacteria in the PC distributed in all the stages. The average age of PC or BPD group were Bacteroidetes (31.96%), and Proteobacteria was patients is relatively close to the onset age of PC or BPD. significantly reduced to 20.85%. At the family and genus Although the HC group has a lower average age, it is still levels, the dominant bacteria of the PC, BPD, and HC groups appropriate for this study as the control considering the were identical, while the corresponded proportion was differ- temporal stability of the oral microbiome. ent from one another. The dominant bacteria in the HC group were Neisseriaceae (28.84%) at the family level and 3.2. Differences in Salivary Microbiota Diversity. To analyze Neisseria (28.60%) at the genus level, while the dominant the bacterial composition and differences among PC, BPD genus in the PC group was Prevotellaceae (23.72%), Neisseria and HC groups, we applied high-throughput sequencing to (12.55%), and Veillonella (12.12%). The changing trend of detect the V3-V4 region of the bacterial 16S rRNA gene. As dominant bacteria in the BPD group relative to the HC group a result, 5,076,996 raw sequence reads with an average length was consistent with that of the PC group, but its magnitude of 445 bps were produced from all the thirty-seven samples. was smaller than that of the PC group. At the species level, In total, 449 OTUs (operational taxonomic units) were the PC, BPD, and HC groups possess the same top five obtained from the sequencing results, which include 19 bacteria, which are Neisseria mucosa, Haemophilus parain- phyla, 35 classes, 70 orders, 103 families, 193 genera, and fluenzae, Prevotella melaninogenica, Veillonella dispar, and 364 species. The Simpson and Heip indexes explicitly showed Fusobacterium periodonticum, but the relative proportions that the differences of oral bacterial community composition of dominant bacteria are different across the three groups. existed among the three groups. The mean value of the Both the dominant bacteria in the HC group (10.97%) and Simpson index was gradually decreased from the HC the BPD group (9.30%) were Neisseria mucosa. In compari- (0.113, SD = 0:039) to the BPD (0.067, SD = 0:023) and then son, the dominant bacteria in the PC group was Veillonella the PC group (0.062, SD = 0:027) (Figure 1(a)). Both PC and dispar (8.32%). Of the top five bacteria, Neisseria mucosa BPD groups were significantly higher than the HC group, and Fusobacterium periodonticum showed great change respectively (P value < 0.05), but there was no obvious differ- amplitude across the HC (10.97%, 2.93%) and PC (4.60%, ence between PC and BPD groups (P value > 0.05). The mean 5.94%) groups. value of the Heip index was gradually increased from the HC We carried out the K-W test and post hoc Dunn’s test of (0.101, SD = 0:028) to the BPD (0.159, SD = 0:041) and then the HC, BPD, and PC groups based on the OTUs at the phy- the PC group (0.178, SD = 0:048) (Figure 1(b)). The Heip lum, family, and species levels. At the phylum level, two of index indicated that the PC group has the highest species uni- nineteen bacterial taxa exhibited a significant difference formity followed by the BPD and HC groups. Moreover, the between groups. The abundance of Proteobacteria in both mean value of Inverse Simpson and Shannon indexes in the the PC and BPD groups was notably lower as compared with PC, BPD and HC groups was increased from the HC the HC group, while the abundance of Spirochaetae in both (10.016, SD = 3:880; 2.970, SD = 0:350) to the BPD (17.019, the PC and BPD groups was markedly higher as compared 4 BioMed Research International

PCoA plot (weighted UniFrac) Community diversity Community evenness 0.24 ⁎⁎ ⁎⁎ 0.2 0.20 ⁎⁎ 0.32 ⁎⁎ 0.16 0.24 0.0 0.12 Heip PC2 (14.3%)

Simpson 0.16 0.08 0.08 –0.2 0.04

0.00 0.00 PC BPD HC PC BPD HC –0.5 0.0 0.5 PC1 (45%) PC PC BPD BPD HC HC (a) (b) (c)

Phylum level Family level Genus level Species level

BPD

0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 1

Proteobacteria Neisseriaceae Neisseria Neisseria mucosa Bacteroidetes Prevotellaceae Haemophilus Haemophilus parainfuenzae Firmicutes Pasteurellaceae Prevotella_7 Prevotella melaninogenica Fusobacteria Streptococcaceae Streptococcus Veillonella dispar Actinobacteria Veillonellaceae Veillonella Fusobacterium periodonticum Saccharibacteria Porphyromonadaceae Porphyromonas Neisseria oralis Spirochaetae Fusobacteriaceae Prevotella Prevotella intermedia Others Burkholderiaceae Fusobacterium Streptococcus salivarius Micrococcaceae Lautropia Neisseria elongata Leptotrichiaceae Alloprevotella Aggregatibacter segnis Others Others Others (d)

Figure 1: Phylogenetic diversity of oral bacteria community composition among the PC, BPD, and HC groups. (a) Histogram depicts community diversity according to the Simpson index among the PC, BPD, and HC groups. (b) Histogram depicts community evenness according to the Heip index among the PC, BPD, and HC groups. (c) Bacterial diversity clustering by weighted UniFrac PCoA of oral bacteria. Each symbol represents a sample and the variance explained by the PCs in parentheses on the axes. (d) Stacked bar depicts the relative proportions of oral bacteria among the PC, BPD, and HC groups at the phylum, family, genus, and species levels. Note: the symbol of “∗” represents the P value between groups which is less than 0.05. with the HC group. At the family level, eight of one hundred the groups (Figure 2). Firstly, we exhibited the oral bacterial and three bacterial taxa showed a significant difference taxa between the HC, BPD, and PC groups (Figure 2(a)). between groups. Neisseriaceae, which was the dominant bac- When the log10 (LDA score) was greater than 3.5, the teria in the HC group, was dramatically lower in the PC dominant bacteria in the PC group were Fusobacterium, group than both the HC and BPD groups. But Prevotellaceae, Megasphaera, Prevotella, Spirochaeta, and Treponema, the which was the dominant microbe in the PC group, displayed dominant bacteria in the HC group were Leptotrichia and no distinct variation between groups. At the species level, Neisseria, and the dominant bacterium in BPD was Seleno- thirty-six of three hundred and sixty-four bacterial taxa monas. We then explored the variant bacterial species showed a distinct difference between groups. However, no between the PC and HC groups (Figure 2(b)). When the statistically significant abundance difference of any bacterial log10 (LDA score) was greater than 4.0, the dominant bacte- taxa was detected between the PC and BPD groups. ria in the PC group were Fusobacterium and Prevotella, even including Fusobacterium periodonticum at the species level. 3.4. Potential Saliva Biomarkers Associated with PC. LEfSe Simultaneously, the dominant bacteria in the HC group were was performed to further uncover the remarkable species of Neisseria and Haemophilus, even including Neisseria mucosa, oral microbiota that characterizes the differences between Haemophilus parainfluenzae, and Leptotrichia goodfellowii at BioMed Research International 5

PC vs. HC vs. BPD

c: g__Megasphaera d: s__Megasphaera_micronuciformis r: p__Spirochaetae s: C__Spirochaetes t: o__Spirochaetales u: f__Spirochaetaceae V: g__Treponema_2 y: g.__Fusobacterium q: s__Bacteroides acidifaciens n: s__Prevotella_pallens w: s__Treponema parvum o: g__Prevotella_6 p: s__unclassifed__g__Prevotella_6 e: p__Proteobacteria f: c__Betaproteobacteria g: o__Neisseriales h: f__Neisseriaceae i: g__Neisseria x: S__Leptotrichia_goodfellowii a: g__Selenomonas_3 j: c__Epsilonproteobacteria k: o__Campylobacterales 1: f_Campylobacteraceae m: g__Campylobacter b: s__unclassifed__g__Selenomonas_3

3.02.52.01.51.00.50.0 5.04.54.03.5 LDA score (log10)

PC HC BPD

(a)

PC vs. HC BPD vs. HC

p__Proteobacteria C_Betaproteobacteria o__Neisseriales f__Neisseriaceae g__Capnocytophaga g__Neisseria p__Saccharibacteria s__Neisseria_mucosa g__Selenomonas_3 s__unclassifed_g_Neisseria c__norank_p__Saccharibacteria C__Gammaproteobacteria f__norank_p__Saccharibacteria o__Pasteurellales g__norank_p__Saccharibacteria f__Pasteurellaceae o__norank_p__Saccharibacteria g__Haemophilus s__unclassifed_g__Selenomonas_3 S__Haemophilus_parainfuenzae_T3T1 c__Epsilonproteobacteria s__Leptotrichia_goodfellowii o__Campylobacterales p__Fusobacteria g__Campylobacter c__Fusobacteriia f__Campylobacteraceae o__Fusobacteriales g__Megasphaera g__Fusobacterium s__Megasphaera_micronuciformis f__Fusobacteriaceae g__revotella_6 S__usobacterium_periodonticum s__unclassifed__g__Prevotella_6 g__Prevotella

5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 LDA score (log10) LDA score (log10)

HC BPD PC HC

(b) (c)

Figure 2: LEfSe and LDA analyses based on OTUs characterize bacteria among the PC, BPD, and HC groups. Cladogram using LEfSe strategy revealing the phylogenetic distribution of oral bacteria related to patients with PC (black), patients with BPD (orange), and HC subjects (green). LDA scores showed a significant bacterial difference between groups. The log10 scaled LDA score is indicated at the bottom, and the red box represents the selected threshold. the species level. Finally, we compared bacterial species nant bacteria in the BPD group were Campylobacter, between the BPD and HC groups (Figure 2(c)). When Capnocytophaga, Megasphaera, Prevotella, and Selenomo- the log10 (LDA score) was greater than 3.5, the domi- nas at the genera level while the dominant bacterium 6 BioMed Research International in the HC group was only Proteobacteria at the phy- is associated with oral bacteria; prior studies showed the lum level. trend that the bacteria of Streptococcus anginosus were elevated [32]. Further, colorectal cancer patients with low 3.5. Specific Changes of the Biological Functions. In order to levels of Fusobacterium nucleatum had a significantly longer characterize the changes in biological functions that may be overall survival time than patients with moderate and high involved in the PC, BPD, and HC groups, the PICRUSt levels of the bacterium [33]. In addition, Streptococcus angi- toolkit was adopted to predict COG functions and KEGG nosus, Streptococcus mitis, and Treponema denticola could pathways. As a result, a total of two hundred and eighteen have a significant role in the carcinogenic process of esopha- KEGG pathways were predicted, and fifty-two were showed geal cancer by causing inflammation [34]. Finally, previous notably different between the three groups based on the investigation described that Oribacterium sp. and Fusobac- number of sequences which are correlated with the pathways. terium sp. could distinguish liver cancer patients from Of the 52 KEGG pathways, 8 showed commonly significant healthy subjects [35]. changes in the comparison between the PC and HC groups as well as the BPD and HC groups (P value < 0.05), and 4. Discussion Figure 3(a) portrayed the value of log10 relative fold change between groups for the eight pathways. Particularly, there PC has a high mortality rate, and the five-year survival rate of were four of the fifty-two pathways involved in the glucose PC patients was about 9% [1]. PC is highly invasive with a metabolism (Figure 3(b)), which are N-glycan biosynthesis, poor prognosis and lacks effective therapeutic drugs [2]. various types of N-glycan biosynthesis, lipopolysaccharide Therefore, prevention at an early stage of PC is of crucial biosynthesis, and peptidoglycan biosynthesis. In addition, importance. As stated above, the oral cavity is characterized forty-two pathways showed significant differences between by the highest microbiome density and possesses the largest the PC and HC groups (P value < 0.05), of which eight path- number of species in the body, and oral bacteria have been ways were upregulated and thirty-two pathways were down- proved associated with many cancers, especially gastro- regulated. After that, an equal number of downregulated intestinal tumors [10]. The architecture of microbiome pathways were selected by the value of log10 relative fold communities orally seems to reflect health status in certain change, and then, the fold change relationship between circumstances, making the analysis of oral microbiomes a groups for the sixteen pathways is described in Figure 3(c). promising approach for cancer diagnostics. Previous studies A total of twenty-five COG function terms were predicted, have shown that the imbalance of oral bacteria community and only five terms exhibited significant difference between composition was associated with the development of PC, the three groups based on the number of sequences which and some bacteria may be the potential diagnostic markers are correlated with the COG terms. The twenty-five COG [17–19]. The role of oral microbiota in the pathogenesis of function terms may be divided into three categories, which cancer may be ascribed to the bacterial stimulation of chronic are the cellular process and signaling, the information storage inflammation or secreting some virulence factors that act in a and processing, and the metabolism [26, 27]. In Figure 4(a), carcinogenic manner. Observably, the association of oral for each of the twenty-five COG function terms, the point bacteria and PC as well as the BPD and the pathogenic mech- diagram displayed the statistical relationships between anism were not yet explored sufficiently. This study is groups, and the stacked bar conveyed the information about expected to contribute to this topic. the number of correlated sequences of each group. The five It is well known that species diversity increased with the COG function items that showed a significant difference rise of species richness as well as uniformity. Herein, we between the three groups are “RNA processing and modifica- explored the species diversity of oral bacterial communities tion,”“cell cycle control, cell division, chromosome partition- among the three groups. The performance of different groups ing,”“inorganic ion transport and metabolism,”“intracellular on the Simpson uncovered that oral bacteria species diversity trafficking, secretion, and vesicular transport,” and “function of both the PC group and the BPD group was obviously unknown.” Except for “function unknown,” Figure 4(b) por- higher than that of the HC group, but there was no significant trayed the distribution of the number of sequences of all the difference between PC and BPD. Other species diversity samples in the three groups. indexes such as Shannon also supported the result and corre- Not only does there exists a strong correlation of oral bac- spond with the previous study [36]. In general, our findings teria with pancreatic cancer, but we also presented here the demonstrated striking changes in the composition of the oral relationship of oral microbiome with other cancers. Firstly, bacterium community among the PC and BPD groups as in contrast to healthy controls, patients with oral squamous compared to the HC group. The differences perhaps related cell carcinoma are usually carrying a higher proportion of indirectly to the biological functions of the pancreas because bacteria of Streptococcus anginosus, Streptococcus mitis, Cap- of the imbalance in the body’s microbiome [18, 36]. nocytophaga gingivalis, Prevotella melaninogenica, and Por- Discovering biomarkers has proven to be the most phyromonas gingivalis orally [28–30]. Then, oral mucosal important and successful way to translate molecular and cancer subjects have higher salivary counts of Streptococcus genomic research output into clinical applications [24]. Our intermedius, Streptococcus constellatus, Streptococcus oralis, efforts systematically shed light on the differences of oral bac- Streptococcus mitis, Streptococcus sanguis, and Streptococcus terial community composition among the PC, BPD, and HC salivarius, which may be treated as the diagnostic indicators groups. While compared with the salivary bacterial commu- [31]. And next, head and neck squamous cell carcinoma also nity of the HC group, the PC group performed the consistent BioMed Research International 7

Pathways commonly changed in PC and BPD 0.6

0.4 PC vs .HC BPD vs. HC PC vs. BPD Various types of N-glycan biosynthesis 0.2 Pathways specially changed in PC Flavonoid biosynthesis N-Glycan biosynthesis Pertussis 0 Protein processing in endoplasmic retic NOD-like receptor signaling pathway Bisphenol degradation

Limonene and pinene degradation Log10 (relative fold change) –0.2 Pyrimidine metabolism

Primary immunodefciency RNA polymerase Amyotrophic lateral sclerosis (ALS) –0.4 Mismatch repair Upregulated (a) Purine metabolism Pathways associated with glycometabolism Peptidoglycan biosynthesis N-Glycan biosynthesis Lipopolysaccharide Homologous recombination ⁎⁎ biosynthesis ⁎ 6000 120000 ⁎ 100000 Systemic lupus erythematosus 4000 80000 Carotenoid biosynthesis 2000 60000 Parkinson’s disease

Cardiac muscle contraction Various types of Peptidoglycan N-glycan biosynthesis biosynthesis Nonhomologous end-joining

Te sequence number 1000 ⁎ 160000 ⁎ Downregulated ⁎⁎ Drug metabolism - cytochrome P450 750 140000 500 Metabolism of xenobiotics by cytochrome P450 120000 250 Retinol metabolism 0 100000 0.0 0.1 0.2 0.3 PC Log10 (relative fold change)

HC PC vs. HC BPD vs. HC BPD PC vs. BPD (b) (c)

Figure 3: The KEGG pathway differences of oral bacteria among the PC, BPD, and HC groups. Statistical significance between groups was calculated by the K-W test and post hoc Dunn’s test (P value < 0.05). (a) Heatmap displayed that eight KEGG categories were detected changed in both PC and BPD groups, and the log10 scaled relative fold change between the groups is indicated at the bottom. (b) Four pathways are associated with glycometabolism and were showed in the box plot based on the sequencing reads. The symbol of “∗” indicates that P value is less than 0.05. (c) Bar plots exhibited sixteen pathways specially changed in the PC group. Thereinto, eight categories are upregulated, whereas eight categories are downregulated. changing trend with the BPD group, but to a larger extent greater the LDA score is, the more significant the microbial than the BPD group. At the phylum level, PC patients tended biomarker is in the comparison. Figure 2 depicts the LEfSe to have higher percentages of Bacteroidetes, Firmicutes, and results. In Figure 2b, Fusobacterium periodonticum was spe- Fusobacteria and lower percentages of Proteobacteria. At cifically abundant in PC patients, meanwhile Neisseria the finer taxonomic levels, we observed differences in the mucosa, Leptotrichia goodfellowii, and Haemophilus parain- abundances of particular genera in PC patients compared fluenzae T3T1 were characteristically enriched in the HC with the HC and BPD groups. For example, the abundance group. However, the population of Leptotrichia goodfellowii of Proteobacteria and its subordinate Neisseriaceae and Neis- in the groups is very low (<0.002%), thus it is not appropriate seria was sharply reduced in the PC group, and the vantage of to treat it as the candidate risk factor for recognizing PC dominant bacteria was destroyed. Meanwhile, the ratio of patients from healthy individuals. As for Haemophilus para- Fusobacteria and its subordinate Fusobacteriaceae, Fusobac- influenzae T3T1, previous research delineated that it was cor- terium, and even the Fusobacterium periodonticum is greatly related with lung cancer, hence it may not be suitable to treat increased in the PC group. it as the especially potential biomarker of PC [37]. As for LEfSe analysis further revealed the differences of the Neisseria mucosa, so far, no research displayed its percentage microbial community between the research groups. The variation in any cancers, and Fusobacterium periodonticum 8 BioMed Research International

RNA processing and modifcation ⁎ 6000

4000 Translation, ribosomal structure and biogenesis 2000 Transcription

Signal transduction mechanisms 0

Secondary metabolite biosynthesis, transport and catabolism ⁎ Cell cycle control, cell division, RNA processing and modifcation chromosome partitioning Replication, recombination and repair 225000 ⁎

Posttranslational modifcation, protein turnover, chaperones 200000

Nucleotide transport and metabolism 175000

Nuclear structure 150000

Lipid transport and metabolism 125000 ⁎ Intracellular trafcking, secretion, and vesicular transport ⁎ Inorganic ion transport and metabolism Inorganic ion transport and metabolism General function prediction only ⁎ ⁎ 9e+05 Function unknown e Te sequence number 8 +05 Extracellular structures e Energy production and conversion 7 +05 e Defense mechanisms 6 +05 e Cytoskeleton 5 +05

Coenzyme transport and metabolism

Chromatin structure and dynamics Intracellular trafcking, secretion, and vesicular transport Cell wall/membrane/envelope biogenesis ⁎ Cell motility 4e+05 ⁎ Cell cycle control, cell division, chromosome partitioning 3e+05 Carbohydrate transport and metabolism

Amino acid transport and metabolism 2e+05

PC PC vs. HC PC vs. BPD BPD vs. HC BPD HC PC vs. BPD HC (a) (b)

Figure 4: The COG function differences of oral bacteria among the PC, BPD, and HC groups. Statistical significance between groups was calculated by the K-W test and post hoc Dunn’s test. (a) For the total twenty-five COG terms, the bar plot depicted the relative changes among the three groups based on the sequence number, and the dot plot exhibited the statistical significance between groups. (b) Box plots showed four COG functions apparently changed between PC and HC groups, which are observably or proven associated with pancreatic cancer. The symbol of “∗” indicates that P value is less than 0.05. also has no evidence of its correlation with cancers except for low-pH products such as lactic acid into weak acids and vol- gastrointestinal tumors [38]. In Figure 2c, Megasphaera atile acids, thereby protecting us from the caries [43]. In addi- micronuciformis performed a high LDA score, but it also tion, the production of acid may contribute to the acidic and appeared in the Figure 2a, which means that no bacteria at hypoxic microenvironment of the tumor, thereby improving the species level can specially be applied for identifying the metastatic capacity [44, 45]. In existing studies, the abun- BPD patients from healthy individuals. In short, such perfor- dance of Neisseria elongata, which belonged to Neisseria mance suggested that the low abundance of Neisseria together with Neisseria mucosa, was significantly lower in mucosa, and the high abundance of Fusobacterium periodon- the PC group than that in the HC group [18]. What we sus- ticum, stands out as potential specific risk factors for PC. pect is that the sharp decrease of Neisseria was usually We are the first to try to comprehensively report the pro- accompanied by an increase in other pathogenic bacteria. tective effect of Neisseria mucosa from PC. Oral diseases such In fact, these studies also described that Toll-like receptors as periodontal disease, caries, and tooth loss have been shown recognize antigens of pathogenic bacteria and thereby modu- to be independent risk factors for the development of PC late immune responses. In addition, inflammatory cytokines [39–41]. However, Neisseria in the oral cavity helps to pre- activate NF-κB-associated cellular pathways and then regu- vent oral diseases [42]. Neisseria in the mouth can metabolize late the expression of genes involved in inflammation, BioMed Research International 9 growth factors, and cell invasion molecules. NF-κB is contin- ation, and survival. The perturbation of the NOD-like recep- uously activated in PC tissue cells. To sum up, the role of tor signaling pathway was confirmed to be significantly Neisseria mucosa in the oral cavity may be a protective factor associated with PC [55], and this pathway also showed for PC, but the association between Neisseria and PC needs noticeable upregulation in our PC group. As bacteria cause further exploration. the occurrence or growth of cancer by inducing an inflamma- In this study, we also first try to comprehensively report tory microenvironment, the inflammatory factor in turn the pathogenetic effect of Fusobacterium periodonticum on leads to cell proliferation, mutations, and oncogene activa- the development of PC. Fusobacterium nucleatum has been tion [11]. Therefore, the alterations of oral bacteria may affect shown to be involved in the development of colon cancer, the development of PC through the natural bacterial antibac- and its concentration levels in the adenomas and cancer tis- terial signaling pathway. sues were higher [46]. Moreover, studies have revealed that The changes in bacteria composition usually accompany FadA, a virulence factor secreted by Fusobacterium nuclea- the alterations of biological functions. We predicted the corre- tum, is an important factor in carcinogenesis [47]. FadA sponding gene functions and metabolic pathways based on the binds to E-catenin, activates β-catenin, leads to upregulation 16S rDNA high-throughput sequencing results via the of “proinflammatory factors, wnt signaling, and oncogenes,” PICRUSt toolkit. Among the predicted results, five COG func- and stimulates cell proliferation [48]. Actually, Fusobacter- tions showed significant association with PC (P value < 0.05). ium periodonticum and Fusobacterium nucleatum are the As for RNA processing and modification, the existing study only bacteria in the genus Fusobacterium that can secrete has proved that the RNA modification of the METTL3 gene FadA virulence factors. In addition, oral bacteria can spread (m6A) promotes the development of PC and antichemothera- to the pancreas through blood circulation and biliary con- py/antiradiation therapy [56]. To intracellular trafficking, duction [15]. A previous study has presented that the detec- secretion, and vesicular transport, it is obvious that the secretion rate of Fusobacterium in PC tissues (283 cases) was tion of insulin from pancreatic β cell is the vital function of the 8.8%, and the presence of Fusobacterium was independently pancreas, and insulin dysregulation is a risk factor for pancre- associated with poor prognosis of PC (P value < 0.05), sug- atic cancer [57]. With regard to inorganic ion transport and gesting that Fusobacterium may be a novel biomarker for metabolism, the previous study showed us that the growth of predicting the prognosis of PC [49]. The case-control studies malignant pancreatic tumors may be related to the metabo- also support our above inference [36, 50]. Taken together, lism of inorganic ion copper, which is essential to participate adhesion of the exogenous virulence factor FadA of the Fuso- in the redox process in cells, and it is also a catalytic cofactor bacterium periodonticum to host epithelial cells is most likely for various enzymes in organisms [58]. All of these reflected one of the molecular mechanisms which lead to PC. the specific association of PC and the above-mentioned bio- Although the pathway we found in the current study is logical functions. predicted based on the 16S sequence, which is not completely true, it can provide some clues for exploring the functions of these bacteria in the carcinogenicity and development of 5. Conclusion pancreatic cancer. In the pathway enrichment outcomes, totally fifty-two KEGG pathways were apparently different In summary, we described the differences of salivary bacte- between the research groups (P value < 0.05). Eight KEGG rium composition among the PC, BPD, and HC groups. The pathways shown in Figure 3(a) were commonly changed in results in this study confirmed that oral bacterial community the PC and BPD groups compared to the HC group. It is composition can be used for distinguishing PC from healthy meaningful that four pathways were related to glucose people as well as the BPD patients. Lower levels of Neisseria metabolism, which contains N-glycan biosynthesis, the syn- mucosa and higher levels of Fusobacterium periodonticum thesis of N-glycan biosynthesis, various types of N-glycan from the oral cavity may be risk factors for the development biosynthesis, lipopolysaccharide biosynthesis, and peptido- of PC. However, more samples are needed to confirm the glycan biosynthesis. Obviously, glucose metabolism has a causal relationship between oral bacteria and PC. Conclu- strong relationship with pancreatic diseases. For example, sively, our findings supplement new knowledge to the medical diabetes usually presented as a complication of chronic pan- problem of early diagnosis or prevention of PC. creatitis [51]. The occurrence and development of tumors are frequently accompanied by the changes of glycosylated proteins, and actually that there are significant changes in carbo- Data Availability hydrate chains on the specific proteins of PC cell [52]. The predictive model based on N-glycan is also proved as an The raw data used to support the findings of this study were effective method for the diagnosis of PC [53]. Previous deposited into the NCBI Sequence Read Archive (SRA) data- research has demonstrated that CA19.9 (carbohydrate anti- base (Bioproject accession Number: SRP237984). gen 19.9) synergizes with oncogene to promote the growth of PC [54]. In addition, among the forty specifically changed KEGG pathways in PC, eight pathways were upregulated and Conflicts of Interest thirty-two pathways were downregulated. Generally speak- ing, the cell signaling pathway usually plays an important The authors declare that there is no conflict of interest role in inflammation, immunity, cell proliferation, differenti- regarding the publication of this article. 10 BioMed Research International

Authors’ Contributions [10] J. A. Aas, B. J. Paster, L. N. Stokes, I. Olsen, and F. E. Dewhirst, “Defining the normal bacterial flora of the oral cavity,” Journal LZ and HW developed the study concept and design. XZ of Clinical Microbiology, vol. 43, no. 11, pp. 5721–5732, 2005. collected the samples. HS and XZ completed the DNA prep- [11] T. M. Karpinski, “Role of oral microbiota in cancer develop- aration and environmental genomic experiments, and HS ment,” Microorganisms, vol. 7, no. 1, p. 20, 2019. performed the data analysis. HS and LZ drafted the manu- [12] J. Ahn, S. Segers, and R. B. Hayes, “Periodontal disease, Por- script. All authors read and approved the final manuscript. phyromonas gingivalis serum antibody levels and orodigestive cancer mortality,” Carcinogenesis, vol. 33, no. 5, pp. 1055– Acknowledgments 1058, 2012. [13] M. Castellarin, R. L. Warren, J. D. Freeman et al., “Fusobacter- The authors thank all the patient volunteers from the Depart- ium nucleatum infection is prevalent in human colorectal car- ment of Hepatobiliary Surgery, the First Affiliated Hospital of cinoma,” Genome Research, vol. 22, no. 2, pp. 299–306, 2012. Army Medical University, and the healthy volunteers that [14] V. Sethi, G. A. Vitiello, D. Saxena, G. Miller, and V. Dudeja, included the Army Medical University staff and Chongqing “The role of the microbiome in immunologic development ” citizens, for their participation in this study. This work was and its implication for pancreatic cancer immunotherapy, – supported by the National Natural Science Foundation of Gastroenterology, vol. 156, no. 7, pp. 2097 2115.e2, 2019. “ China (No. 31771468). [15] I. Brook and E. H. Frazier, Microbiological analysis of pancreatic abscess,” Clinical Infectious Diseases, vol. 22, no. 2, pp. 384-385, 1996. Supplementary Materials [16] J. H. Meurman, “Oral microbiota and cancer,” Journal of Oral Microbiology, vol. 2, no. 1, 2010. Table S1: clinical characteristics for all volunteers. Figure S1: bar plots depicted oral bacteria-alpha-diversity differences among [17] X. Fan, E. M. Fletcher, S. M. Gibbons, M. Bouvet, K. S. Doran, and S. T. Kelley, “Human oral microbiome and prospective the PC, BPD, and HC groups. (Supplementary Materials) risk for pancreatic cancer: a population-based nested case- control study,” Gut, vol. 67, no. 1, pp. 120–127, 2017. References [18] J. J. Farrell, L. Zhang, H. Zhou et al., “Variations of oral microbiota are associated with pancreatic diseases including pancre- [1] F. Bray, J. Ferlay, I. Soerjomataram, R. L. Siegel, L. A. Torre, atic cancer,” Gut, vol. 61, no. 4, pp. 582–588, 2012. and A. Jemal, “Global cancer statistics 2018: GLOBOCAN esti- mates of incidence and mortality worldwide for 36 cancers in [19] P. J. Torres, E. M. Fletcher, S. M. Gibbons, M. Bouvet, K. S. “ 185 countries,” CA: a Cancer Journal for Clinicians, vol. 68, Doran, and S. T. Kelley, Characterization of the salivary ” no. 6, pp. 394–424, 2018. microbiome in patients with pancreatic cancer, PeerJ, vol. 3, [2] J. Kleeff, M. Korc, M. Apte et al., “Pancreatic cancer,” Nature article e1373, 2015. “ Reviews Disease Primers, vol. 2, no. 1, article 16022, 2016. [20] T. Magoc and S. L. Salzberg, FLASH: fast length adjustment of “ short reads to improve genome assemblies,” Bioinformatics, [3] R. L. Siegel, K. D. Miller, and A. Jemal, Cancer statistics, – 2019,” CA: A Cancer Journal for Clinicians, vol. 69, no. 1, vol. 27, no. 21, pp. 2957 2963, 2011. pp. 7–34, 2019. [21] J. Kuczynski, J. Stombaugh, W. A. Walters, A. Gonzalez, J. G. “ [4] M. S. De La Cruz, A. P. Young, and M. T. Ruffin, “Diagnosis Caporaso, and R. Knight, Using QIIME to analyze 16S rRNA ” and management of pancreatic cancer,” American Family Phy- gene sequences from microbial communities, Current Proto- sician, vol. 89, no. 8, pp. 626–632, 2014. cols in Bioinformatics, vol. 36, no. 1, 2011. “ [5] C. J. Yeo, J. L. Cameron, K. D. Lillemoe et al., “Pancreaticoduo- [22] J. R. Cole, B. Chai, T. L. Marsh et al., The Ribosomal Database denectomy for cancer of the head of the pancreas. 201 Project (RDP-II): previewing a new autoaligner that allows ” patients,” Annals of Surgery, vol. 221, no. 6, pp. 721–733, regular updates and the new prokaryotic taxonomy, Nucleic 1995, discussion 731-723. Acids Research, vol. 31, no. 1, pp. 442-443, 2003. “ [6] A. Frena, “SPan-1 and exocrine pancreatic carcinoma. The [23] P. Virtanen, R. Gommers, T. E. Oliphant et al., SciPy 1.0: fun- fi ” clinical role of a new tumor marker,” The International Journal damental algorithms for scienti c computing in Python, – of Biological Markers, vol. 16, no. 3, pp. 189–197, 2001. Nature Methods, vol. 17, no. 3, pp. 261 272, 2020. “ [7] J. E. Kim, K. T. Lee, J. K. Lee, S. W. Paik, J. C. Rhee, and K. W. [24] N. Segata, J. Izard, L. Waldron et al., Metagenomic biomarker ” Choi, “Clinical usefulness of carbohydrate antigen 19-9 as a discovery and explanation, Genome Biology, vol. 12, no. 6, screening test for pancreatic cancer in an asymptomatic popu- article R60, 2011. lation,” Journal of Gastroenterology and Hepatology, vol. 19, [25] M. G. I. Langille, J. Zaneveld, J. G. Caporaso et al., “Predictive no. 2, pp. 182–186, 2004. functional profiling of microbial communities using 16S rRNA [8] Y. Zhang, J. Yang, H. Li, Y. Wu, H. Zhang, and W. Chen, marker gene sequences,” Nature Biotechnology, vol. 31, no. 9, – “Tumor markers CA19-9, CA242 and CEA in the diagnosis pp. 814 821, 2013. of pancreatic cancer: a meta-analysis,” International Journal [26] S. V. Angiuoli, M. Matalka, A. Gussman et al., “CloVR: a vir- of Clinical and Experimental Medicine, vol. 8, no. 7, tual machine for automated and portable sequence analysis pp. 11683–11691, 2015. from the desktop using cloud computing,” BMC Bioinformat- [9] The Integrative HMP (iHMP) Research Network Consortium, ics, vol. 12, no. 1, p. 356, 2011. “The Integrative Human Microbiome Project: dynamic analy- [27] M. Y. Galperin, K. S. Makarova, Y. I. Wolf, and E. V. Koonin, sis of microbiome-host omics profiles during periods of “Expanded microbial genome coverage and improved protein human health and disease,” Cell Host & Microbe, vol. 16, family annotation in the COG database,” Nucleic Acids no. 3, pp. 276–289, 2014. Research, vol. 43, no. D1, pp. D261–D269, 2015. BioMed Research International 11

[28] L. V. Galvão-Moreira and M. C. F. N. da Cruz, “Oral micro- [44] S. J. Lunt, N. Chaudary, and R. P. Hill, “The tumor microenvi- biome, periodontitis and risk of head and neck cancer,” Oral ronment and metastatic disease,” Clinical & Experimental Oncology, vol. 53, pp. 17–19, 2016. Metastasis, vol. 26, no. 1, pp. 19–34, 2009. [29] D. L. Mager, A. D. Haffajee,P.M.Devlin,C.M.Norris,M.R. [45] E. A. Mazzio, B. Smith, and K. F. Soliman, “Evaluation of Posner, and J. M. Goodson, “The salivary microbiota as a diag- endogenous acidic metabolic products associated with carbo- nostic indicator of oral cancer: a descriptive, non-randomized hydrate metabolism in tumor cells,” Cell Biology and Toxicol- study of cancer-free and oral squamous cell carcinoma subjects,” ogy, vol. 26, no. 3, pp. 177–188, 2010. Journal of Translational Medicine,vol.3,no.1,p.27,2005. [46] M. R. Rubinstein, X. Wang, W. Liu, Y. Hao, G. Cai, and Y. W. [30] M. Sasaki, C. Yamaura, Y. Ohara-Nemoto et al., “Streptococ- Han, “Fusobacterium nucleatum promotes colorectal carcinogen- cus anginosus infection in oral cancer and its infection route,” esis by modulating E-cadherin/β-catenin signaling via its FadA Oral Diseases, vol. 11, no. 3, pp. 151–156, 2005. adhesin,” Cell Host & Microbe, vol. 14, no. 2, pp. 195–206, 2013. [31] H. Sakamoto, H. Naito, Y. Ohta et al., “Isolation of bacteria [47] Y. W. Han, A. Ikegami, C. Rajanna et al., “Identification and from cervical lymph nodes in patients with oral cancer,” characterization of a novel adhesin unique to oral fusobac- Archives of Oral Biology, vol. 44, no. 10, pp. 789–793, 1999. teria,” Journal of Bacteriology, vol. 187, no. 15, pp. 5330– [32] K. Shiga, M. Tateda, S. Saijo et al., “Presence of Streptococcus 5340, 2005. infection in extra-oropharyngeal head and neck squamous cell [48] P. Gholizadeh, H. Eslami, and H. S. Kafil, “Carcinogenesis carcinoma and its implication in carcinogenesis,” Oncology mechanisms of Fusobacterium nucleatum,” Biomedicine & Reports, vol. 8, no. 2, pp. 245–248, 2001. Pharmacotherapy, vol. 89, pp. 918–925, 2017. [33] L. Flanagan, J. Schmid, M. Ebert et al., “Fusobacterium nucle- [49] K. Mitsuhashi, K. Nosho, Y. Sukawa et al., “Association of atum associates with stages of colorectal neoplasia develop- Fusobacterium species in pancreatic cancer tissues with molec- ment, colorectal cancer and disease outcome,” European ular features and prognosis,” Oncotarget, vol. 6, no. 9, Journal of Clinical Microbiology & Infectious Diseases, pp. 7209–7220, 2015. – vol. 33, no. 8, pp. 1381 1390, 2014. [50] E. del Castillo, R. Meier, M. Chung et al., “The microbiomes of [34] M. Narikiyo, C. Tanabe, Y. Yamada et al., “Frequent and pref- pancreatic and duodenum tissue overlap and are highly subject erential infection of Treponema denticola, Streptococcus specific but differ between pancreatic cancer and noncancer mitis, and Streptococcus anginosus in esophageal cancers,” subjects,” Cancer Epidemiology, Biomarkers & Prevention, Cancer Science, vol. 95, no. 7, pp. 569–574, 2004. vol. 28, no. 2, pp. 370–383, 2019. [35] H. Lu, Z. Ren, A. Li et al., “Deep sequencing reveals microbiota [51] Y. Liu, D. Wang, H. L. Guo et al., “Risk factors and nomogram dysbiosis of tongue coat in patients with liver carcinoma,” Sci- for diabetes mellitus in idiopathic chronic pancreatitis,” Jour- entific Reports, vol. 6, no. 1, article 33142, 2016. nal of Gastroenterology and Hepatology, vol. 35, no. 2, [36] H. Lu, Z. Ren, A. Li et al., “Tongue coating microbiome data pp. 343–352, 2019. distinguish patients with pancreatic head cancer from healthy [52] T. Yue, I. J. Goldstein, M. A. Hollingsworth, K. Kaul, R. E. controls,” Journal of Oral Microbiology, vol. 11, no. 1, article Brand, and B. B. Haab, “The prevalence and nature of glycan 1563409, 2019. alterations on specific proteins in pancreatic cancer patients [37] U. Kosikowska, A. Biernasiuk, P. Rybojad, R. Los, and revealed using antibody-lectin sandwich arrays,” Molecular & A. Malm, “Haemophilus parainfluenzae as a marker of the Cellular Proteomics, vol. 8, no. 7, pp. 1697–1707, 2009. upper respiratory tract microbiota changes under the influence [53] Y. P. Zhao, P. T. Zhou, W. P. Ji et al., “Validation of N-glycan of preoperative prophylaxis with or without postoperative markers that improve the performance of CA19-9 in pancre- treatment in patients with lung cancer,” BMC Microbiology, atic cancer,” Clinical and Experimental Medicine, vol. 17, vol. 16, no. 1, p. 62, 2016. no. 1, pp. 9–18, 2017. [38] A. Saffarian, C. Mulet, B. Regnault et al., “Crypt- and mucosa- [54] D. D. Engle, H. Tiriac, K. D. Rivera et al., “The glycan CA19-9 associated core microbiotas in humans and their alteration in promotes pancreatitis and pancreatic cancer in mice,” Science, colon cancer patients,” mBio, vol. 10, no. 4, 2019. vol. 364, no. 6446, pp. 1156–1162, 2019. [39] P. P. Hujoel, M. Drangsholt, C. Spiekerman, and N. S. Weiss, [55] A. Miskiewicz, G. Szparecki, M. Durlik, G. Rydzewska, “An exploration of the periodontitis-cancer association,” I. Ziobrowski, and R. Gorska, “The Q705K and F359L single- Annals of Epidemiology, vol. 13, no. 5, pp. 312–316, 2003. nucleotide polymorphisms of NOD-like receptor signaling [40] D. S. Michaud, K. Joshipura, E. Giovannucci, and C. S. Fuchs, pathway: association with chronic pancreatitis, pancreatic can- “A prospective study of periodontal disease and pancreatic cer, and periodontitis,” Archivum Immunologiae et Therapiae cancer in US male health professionals,” Journal of the Experimentalis (Warsz), vol. 63, no. 6, pp. 485–494, 2015. National Cancer Institute, vol. 99, no. 2, pp. 171–175, 2007. [56] K. Taketo, M. Konno, A. Asai et al., “The epitranscriptome [41] R. Z. Stolzenberg-Solomon, K. W. Dodd, M. J. Blaser, m6A writer METTL3 promotes chemo- and radioresistance J. Virtamo, P. R. Taylor, and D. Albanes, “Tooth loss, pancre- in pancreatic cancer cells,” International Journal of Oncology, atic cancer, and Helicobacter pylori,” The American Journal of vol. 52, no. 2, pp. 621–629, 2018. Clinical Nutrition, vol. 78, no. 1, pp. 176–181, 2003. [57] B. M. Wolpin, Y. Bao, Z. R. Qian et al., “Hyperglycemia, insu- [42] G. Liu, C. M. Tang, and R. M. Exley, “Non-pathogenic Neis- lin resistance, impaired pancreatic β-cell function, and risk of seria: members of an abundant, multi-habitat, diverse genus,” pancreatic cancer,” Journal of the National Cancer Institute, Microbiology, vol. 161, no. 7, pp. 1297–1312, 2015. vol. 105, no. 14, pp. 1027–1035, 2013. [43] E. Hoshino and A. Araya, “Lactate degradation by [58] Z. Yu, R. Zhou, Y. Zhao et al., “Blockage of SLC31A1- polysaccharide-producing Neisseria isolated from human den- dependent copper absorption increases pancreatic cancer cell tal plaque,” Archives of Oral Biology, vol. 25, no. 3, pp. 211-212, autophagy to resist cell death,” Cell Proliferation, vol. 52, 1980. no. 2, article e12568, 2019.